This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: make grokfield handle locations


2008/8/19 Aldy Hernandez <aldyh@redhat.com>:
>
>> >         struct { int x; int; } a;
>> >                            /* { dg-warning "does not declare anything" }
>>
>> That seems highly incompatible.  Instead, for now I advise modifying
>> testcases to move the single token in question onto its own line so
>> diagnostics can be tested to be associated with it.
>
> Can we talk about how best to fix the dg/ machinery?  I don't mind
> fixing it once and for all.  I'd like something that's intuitive and
> easy for folks to use, but obviously, what I thought was intuitive is not,
> so I'm open for suggestions.

Fixing the dg machinery to work with column numbers would be great and
a huge step forward.

If you have a proposal, you have to explain it in full and consider
its consequences. What Joseph means with "highly incompatible" is that
your proposed syntax will break all current tests. It is simply not
feasible for any human being to update the whole testsuite by hand.
So any new syntax has to be compatible with the current tests or the
current test could be updated to the new syntax mostly in an automatic
way. I don't see how that can be done with your approach.

I would also like something that is intuitive but we are very
constrained by the limits of DejaGNU and the already existing tests. I
think that my proposal takes into account those.

Cheers,

Manuel.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]